基于生成式对抗网络的路网交通流数据补全方法

交通信息的完整性直接影响着城市交通管理的效率.针对城市道路交通中因路段检测器覆盖不全或设备损坏等造成的流量检测数据缺失问题，本文提出基于生成式对抗网络 (Generative Adversarial Network，GAN)算法的交通流量数据补全方法.首先，以路段实际流量为基础，进行图像化处理生成路网二维信息图；其次，计算考虑时空信息补偿的路网关联矩阵，利用GAN算法分析并实现路网二维信息图缺失部分的补全，进而得到路段交通流量的完整数据；最后，利用实际数据，对比分析了本文方法与相空间重构的卡尔曼滤波方法对缺失数据的补全情况.实例分析结果验证了本文方法的可行性和有效性.

Urban Traffic Flow Data Recovery Method Based on Generative Adversarial Network

The completeness of traffic information has a direct influence on urban traffic management efficiency. Aiming at the problem of missing detection data in urban road traffic caused by incomplete coverage of road detectors or equipment damage, this paper proposes a method of traffic flow data completion based on Generative Adversarial Network (GAN) algorithm. Firstly, the traffic flow data of the links are processed graphically to generate the two-dimensional information map of network. Secondly, the road network association matrix which considered the temporal and spatial information compensation is calculated. The GAN algorithm is used to analyze and archive the completion and reconstruction the missing part of information map, and then the traffic flow complete data of the link is obtained. Finally, the proposed method is compared and analyzed with Kalman filtering method of phase space reconstruction. And the results show that the proposed method in this paper is effective.